i'm yajat, a software engineer focused on building scalable systems and practical ai

currently at Eurofins IT Solutions, Bangalore

notable projects

Built, shipped, and running in production.

AIComputer VisionDistributed Systems 2026

Automated Fish Measurement System

End-to-end computer vision platform that replaced manual fish measurement workflows for enterprise clients. Led architecture from hackathon prototype to full production deployment on Azure with Temporal orchestration, GPU-backed ML endpoints, and horizontally scalable worker pods processing ~960 images daily.

95%

Processing time cut

960

Images per day

15 min

Was 6-10 hours

Full-StackSelf-HostedSvelte 2021

Gift Inventory Management System

Full-stack inventory system built with Svelte, Express.js, and MongoDB, serving 500+ active users with role-based access control and modular APIs. Self-hosted on NGINX with PM2 workers and maintained in continuous production for 4+ years with uptime, version updates, and backward compatibility.

500+

Active users

4+ yrs

In production

80%

Effort reduced

experience

From concept to production, end-to-end.

Associate Software Developer

Eurofins IT Solutions

AI Engineering Team

Led end-to-end architecture and implementation of automated fish measurement system from concept to production
Built initial prototype during internal hackathon and drove it to full enterprise deployment
Reduced manual batch processing time from 6–10 hours to 15 minutes per 80 images
Designed Angular-based upload workflow with controlled concurrency and UI feedback states
Introduced and productionized Temporal for asynchronous workflow orchestration with horizontally scalable worker pods on AKS
Integrated GPU-backed Azure ML endpoints, addressed endpoint throttling, tuned CPU parameters, and scaled instances for better throughput
Implemented AprilTag-based perspective correction and morphological calibration logic
Built skeletonization pipeline using medial axis algorithm to measure fish from SAM2 masks
Validated output on 100+ annotated samples within 10% of manual measurement
Designed blob caching strategy to prevent redundant downloads across distributed workers
Presented product directly to US-based enterprise clients; system processes ~960 images/day

Archiving Team

Contributed to distributed archival system storing study reports with 30–40 year retention
Built custom .NET streaming uploader handling files up to 10 GB
Implemented checksum validation for file integrity and version control
Integrated Azure Service Bus queues with dead letter handling
Re-architected email notification flow with database-backed state tracking; reduced latency from 24 hours to 5 minutes

Engineering Contributions

Introduced Tailwind CSS to frontend stack, led Angular upgrade from v18 to v21
Influenced adoption of Temporal for structured async workflows across teams
Owned feature delivery end-to-end: design, development, testing, UAT coordination, rollout, and documentation
Worked in Agile, PR-reviewed development model with zero-defect quality expectations

Feb 2024 – Present

Technical Cloud Consultant

Oracle

Built OCR-based invoice extraction solution using Python and OCI ML services
Developed VBCS interface enabling invoice upload and structured JSON output for ERP ingestion
Onboarded 4 enterprise clients including Berger Paints and Maersk
Mentored 7 junior consultants over 4 months
Conducted 2-day workshop for 30+ participants on Python and Generative AI APIs
Built internal data transformation tools using Python and Pandas for ODI ingestion pipelines